CAT-Net: A Cross-Slice Attention Transformer Model for Prostate Zonal Segmentation in MRI
Prostate cancer is the second leading cause of cancer death among men in the United States. The diagnosis of prostate MRI often relies on the accurate prostate zonal segmentation. However, state-of-the-art automatic segmentation methods often fail to produce well-contained volumetric segmentation of the prostate zones since certain slices of prostate MRI, such as base and apex slices, are harder to segment than other slices. This difficulty can be overcome by accounting for the cross-slice relationship of adjacent slices, but current methods do not fully learn and exploit such relationships. In this paper, we propose a novel cross-slice attention mechanism, which we use in a Transformer module to systematically learn the cross-slice relationship at different scales. The module can be utilized in any existing learning-based segmentation framework with skip connections. Experiments show that our cross-slice attention is able to capture the cross-slice information in prostate zonal segmentation and improve the performance of current state-of-the-art methods. Our method improves segmentation accuracy in the peripheral zone, such that the segmentation results are consistent across all the prostate slices (apex, mid-gland, and base).